Talk:Doomsday argument

Obviously this has some room for improvement. For instance, I'm drawing a blank on categories. Anyone care to add some?

My personal take on the argument is very simple: improbable things happen. Wehpudicabok [話] [変] [留] 09:34, 8 April 2014 (UTC)

I'm no mathematician, but production of tanks is theoretically linear, since tanks can't produce more of themselves, and people can, so our curve would be logarithmic. This reminds me of Thomas Malthus. LittleShortForAStormtrooper (talk) 07:20, 12 April 2014 (UTC)

By the same logic, if someone examines a serial number out of the first five tanks produced of a billion-tank production run, they can conclude there will only be about 10 tanks. Dubious (talk) 16:39, 13 June 2017 (UTC)

The chances of them encountering one of the first five tanks is miniscule though. Christopher (talk) 16:51, 13 June 2017 (UTC)

Dubious, when you're taking the sample you don't know how long the production of tanks will last. You're just taking a random sample of serial numbers. If you get tank #3, you could be in the very first batch of what will eventually be 8,800 Panzer IVs. Or, you could have tank #3 of the 5 Neubaufahrzeug built. Since you're taking your sample randomly, why should you assume you just happened to get the first few? -- Onychoprion (talk) 17:30, 13 June 2017 (UTC)

You don't know; that's the entire point. Your analysis can only estimate something about the existing population. If that population is 5 tanks—then you can estimate only something about 5 tanks. This is exactly the WW2 case; it could be estimated how many tanks there were. It says nothing about future tanks, or where we are in the production line of tanks, etc. Similar, say someone starts writing down a bunch of numbers sequentially; at some arbitrary point, you take a sample of the numbers written down. You can similarly very roughly estimate how many numbers have been written down... but this says nothing about how long the person will continue to write, or at what speed. The sample of a population at a point is not the same as a sample of a point. Dubious (talk) 21:22, 13 June 2017 (UTC)

Though you aren't a random sample of humans who have so far existed. You're one of the last humans that have so far existed, the equivalent of discovering tank 5 has just rolled off the production line. You know there were around 5 tanks made. You are a random number of all humans who will ever exist, because you could've just as easily been born in the year A.D. 3, or 20,000 years ago, or 20,000 years from now (assuming humans are still around then).

This page has a better primer than what we have (I should add it as a See Also). To steal from him and extend the tank analogy, say someone gives you a tank of a particular model, and says that there are some models that will have 10,000 units built, and some that will only have 10 units built. You don't know which your model is, but your tank does have a serial number: 7. It doesn't mean your model won't have 10,000 unit, but the fact that you got a 7 when the statistically average serial number for a 10,000-unit-model has a 5-digit serial, is significant information, whereas a 7 is what you'd expect if picking randomly from 10.

Basically, your argument boils down to "But we could be in the first 5% of humanity!" and yes, we could, but it's significantly more likely we're somewhere in the middle 90%, because most humans who will ever exist will be. -- Onychoprion (talk) 23:41, 13 June 2017 (UTC)

No, that's not the argument. The argument is that this is a fallacy; it's a subtle statistical equivocation and non-sequitur. There are numerous problems with the argument, which is why people likely "disagree" on why: an example of fractal wrongness that's not just lunacy. But, how about an analogy that makes the non-sequitur a bit more obvious, without the equivocation:

You get on a bus. You don't know how long the trip will be, but you can see people getting on the bus and off the bus. After three minutes, you count five people on the bus. You figure it's a safe bet the bus holds at least 10 people. Therefore the trip is probably about 6 minutes.

The main point here is that taking a sample of the population now is not the same as taking a random sample of the entire population for all time. It's equivocating two entirely different populations via "a random sample". Taking a sample of all existing tanks and saying "this is more likely in the middle 90% than the outlying 10%" works because you have the whole population, or at least, you're only making an estimate about the existing population. Critically, just picking "the latest one" is not a random sample of the population for all time, and therefore we can't say it's more likely to be in the middle 90%.

This is just scraping the surface of a primary statistical issue. Really, there are so many things wrong with the argument and its assumptions, it's actually more interesting for that than anything else. Dubious (talk) 16:12, 16 June 2017 (UTC)